Xata Ditches OLAP Stack: Vanilla Postgres as Data Warehouse

industry sources is tracking a notable shift in how teams approach data warehouse infrastructure. Xata abandoned the traditional OLAP playbook - no Snowflake, no BigQuery, no ClickHouse - and instead built their product analytics system directly on vanilla Postgres. The stack relies on three core mechanics: materialized views for pre-aggregated data, pg_cron for scheduled refresh jobs, and copy-on-write branches for isolated query environments.

This isn't a marketing move. Xata runs their own platform on this infrastructure and publishes real performance numbers. They're storing event data, running analytical queries against it, and serving results from the same Postgres instance their operational data lives on. The materialized views handle the heavy lifting - pre-computing aggregations so queries don't scan raw events every time.

The copy-on-write branches deserve attention here. Instead of creating expensive full clones for analytical queries, branches point to shared data until writes occur. This reduces storage overhead and lets analytics queries run without blocking transactional workloads.

• Materialized views pre-aggregate event data and reduce per-query scan time
• pg_cron handles automated refresh schedules without external orchestration
• Copy-on-write branches isolate analytical workloads while sharing underlying data
• Single Postgres instance eliminates ETL sync complexity between operational and analytical layers

Traditional data warehouses optimize for scale but introduce operational friction. You need ETL pipelines to move data, separate storage infrastructure, identity and access management across systems, and teams trained on vendor-specific SQL dialects. Xata's approach eliminates that middle layer.

The practical win: if your analytical queries can tolerate materialized view refresh intervals (typically seconds to minutes), Postgres handles the workload. You keep data in one place, run analytics with standard SQL, and avoid the networking overhead of cross-system queries. The constraint is query latency during refresh windows and storage limits on a single machine.

This approach scales to a real threshold - roughly when your event volume exceeds what one Postgres instance can handle with reasonable refresh times. For product analytics at most SaaS scale, that threshold is higher than you'd expect. Xata published their numbers: they're running petabyte-scale analytics on this pattern.

The operational reality: you're trading managed scaling (Snowflake/BigQuery) for lower operational overhead and tighter integration with your application database.

• Eliminates ETL pipeline maintenance and data sync complexity
• Single source of truth reduces schema drift and identity mismatches
• Query patterns must fit materialized view refresh cadence, not real-time streaming
• Storage and compute scale together on one machine - hit limits sooner than cloud warehouses

Postgres continues eating up use cases that seemed locked into specialized tools. The ecosystem matured: Postgres Foreign Data Wrappers handle cross-system joins, JSON operators handle unstructured data, and now materialized views plus scheduling handle warehouse patterns. For teams that don't need petabyte scale or sub-second analytics refresh, the simplification is real.

This doesn't mean OLAP databases are obsolete. Snowflake, BigQuery, and ClickHouse still own teams with massive query volume, strict latency requirements, or complex multi-tenant analytics. But the decision framework shifts. The baseline is now 'can we do this in Postgres' rather than 'we need a data warehouse.' Builders should flip their defaults.

Xata's competitive move is clear: they're positioning Postgres-as-warehouse as a simpler, more integrated alternative to traditional stacks. If it works for their own analytics, it becomes a credible story for customers. This accelerates the window where teams actually evaluate whether they need separate infrastructure.

The momentum in this space continues to accelerate.

• Postgres ecosystem matured enough to handle warehouse workloads for mid-market analytics
• Simplification and integration beat raw scale in decision-making for teams under 10B events/month
• Copy-on-write isolation patterns are becoming table stakes for Postgres-native analytics